1 | // SPDX-License-Identifier: GPL-2.0-or-later |
2 | /* |
3 | * Glue code for SHA-256 implementation for SPE instructions (PPC) |
4 | * |
5 | * Based on generic implementation. The assembler module takes care |
6 | * about the SPE registers so it can run from interrupt context. |
7 | * |
8 | * Copyright (c) 2015 Markus Stockhausen <stockhausen@collogia.de> |
9 | */ |
10 | |
11 | #include <crypto/internal/hash.h> |
12 | #include <linux/init.h> |
13 | #include <linux/module.h> |
14 | #include <linux/mm.h> |
15 | #include <linux/types.h> |
16 | #include <crypto/sha2.h> |
17 | #include <crypto/sha256_base.h> |
18 | #include <asm/byteorder.h> |
19 | #include <asm/switch_to.h> |
20 | #include <linux/hardirq.h> |
21 | |
22 | /* |
23 | * MAX_BYTES defines the number of bytes that are allowed to be processed |
24 | * between preempt_disable() and preempt_enable(). SHA256 takes ~2,000 |
25 | * operations per 64 bytes. e500 cores can issue two arithmetic instructions |
26 | * per clock cycle using one 32/64 bit unit (SU1) and one 32 bit unit (SU2). |
27 | * Thus 1KB of input data will need an estimated maximum of 18,000 cycles. |
28 | * Headroom for cache misses included. Even with the low end model clocked |
29 | * at 667 MHz this equals to a critical time window of less than 27us. |
30 | * |
31 | */ |
32 | #define MAX_BYTES 1024 |
33 | |
34 | extern void ppc_spe_sha256_transform(u32 *state, const u8 *src, u32 blocks); |
35 | |
36 | static void spe_begin(void) |
37 | { |
38 | /* We just start SPE operations and will save SPE registers later. */ |
39 | preempt_disable(); |
40 | enable_kernel_spe(); |
41 | } |
42 | |
43 | static void spe_end(void) |
44 | { |
45 | disable_kernel_spe(); |
46 | /* reenable preemption */ |
47 | preempt_enable(); |
48 | } |
49 | |
50 | static inline void ppc_sha256_clear_context(struct sha256_state *sctx) |
51 | { |
52 | int count = sizeof(struct sha256_state) >> 2; |
53 | u32 *ptr = (u32 *)sctx; |
54 | |
55 | /* make sure we can clear the fast way */ |
56 | BUILD_BUG_ON(sizeof(struct sha256_state) % 4); |
57 | do { *ptr++ = 0; } while (--count); |
58 | } |
59 | |
60 | static int ppc_spe_sha256_update(struct shash_desc *desc, const u8 *data, |
61 | unsigned int len) |
62 | { |
63 | struct sha256_state *sctx = shash_desc_ctx(desc); |
64 | const unsigned int offset = sctx->count & 0x3f; |
65 | const unsigned int avail = 64 - offset; |
66 | unsigned int bytes; |
67 | const u8 *src = data; |
68 | |
69 | if (avail > len) { |
70 | sctx->count += len; |
71 | memcpy((char *)sctx->buf + offset, src, len); |
72 | return 0; |
73 | } |
74 | |
75 | sctx->count += len; |
76 | |
77 | if (offset) { |
78 | memcpy((char *)sctx->buf + offset, src, avail); |
79 | |
80 | spe_begin(); |
81 | ppc_spe_sha256_transform(state: sctx->state, src: (const u8 *)sctx->buf, blocks: 1); |
82 | spe_end(); |
83 | |
84 | len -= avail; |
85 | src += avail; |
86 | } |
87 | |
88 | while (len > 63) { |
89 | /* cut input data into smaller blocks */ |
90 | bytes = (len > MAX_BYTES) ? MAX_BYTES : len; |
91 | bytes = bytes & ~0x3f; |
92 | |
93 | spe_begin(); |
94 | ppc_spe_sha256_transform(state: sctx->state, src, blocks: bytes >> 6); |
95 | spe_end(); |
96 | |
97 | src += bytes; |
98 | len -= bytes; |
99 | } |
100 | |
101 | memcpy((char *)sctx->buf, src, len); |
102 | return 0; |
103 | } |
104 | |
105 | static int ppc_spe_sha256_final(struct shash_desc *desc, u8 *out) |
106 | { |
107 | struct sha256_state *sctx = shash_desc_ctx(desc); |
108 | const unsigned int offset = sctx->count & 0x3f; |
109 | char *p = (char *)sctx->buf + offset; |
110 | int padlen; |
111 | __be64 *pbits = (__be64 *)(((char *)&sctx->buf) + 56); |
112 | __be32 *dst = (__be32 *)out; |
113 | |
114 | padlen = 55 - offset; |
115 | *p++ = 0x80; |
116 | |
117 | spe_begin(); |
118 | |
119 | if (padlen < 0) { |
120 | memset(p, 0x00, padlen + sizeof (u64)); |
121 | ppc_spe_sha256_transform(state: sctx->state, src: sctx->buf, blocks: 1); |
122 | p = (char *)sctx->buf; |
123 | padlen = 56; |
124 | } |
125 | |
126 | memset(p, 0, padlen); |
127 | *pbits = cpu_to_be64(sctx->count << 3); |
128 | ppc_spe_sha256_transform(state: sctx->state, src: sctx->buf, blocks: 1); |
129 | |
130 | spe_end(); |
131 | |
132 | dst[0] = cpu_to_be32(sctx->state[0]); |
133 | dst[1] = cpu_to_be32(sctx->state[1]); |
134 | dst[2] = cpu_to_be32(sctx->state[2]); |
135 | dst[3] = cpu_to_be32(sctx->state[3]); |
136 | dst[4] = cpu_to_be32(sctx->state[4]); |
137 | dst[5] = cpu_to_be32(sctx->state[5]); |
138 | dst[6] = cpu_to_be32(sctx->state[6]); |
139 | dst[7] = cpu_to_be32(sctx->state[7]); |
140 | |
141 | ppc_sha256_clear_context(sctx); |
142 | return 0; |
143 | } |
144 | |
145 | static int ppc_spe_sha224_final(struct shash_desc *desc, u8 *out) |
146 | { |
147 | __be32 D[SHA256_DIGEST_SIZE >> 2]; |
148 | __be32 *dst = (__be32 *)out; |
149 | |
150 | ppc_spe_sha256_final(desc, out: (u8 *)D); |
151 | |
152 | /* avoid bytewise memcpy */ |
153 | dst[0] = D[0]; |
154 | dst[1] = D[1]; |
155 | dst[2] = D[2]; |
156 | dst[3] = D[3]; |
157 | dst[4] = D[4]; |
158 | dst[5] = D[5]; |
159 | dst[6] = D[6]; |
160 | |
161 | /* clear sensitive data */ |
162 | memzero_explicit(s: D, SHA256_DIGEST_SIZE); |
163 | return 0; |
164 | } |
165 | |
166 | static int ppc_spe_sha256_export(struct shash_desc *desc, void *out) |
167 | { |
168 | struct sha256_state *sctx = shash_desc_ctx(desc); |
169 | |
170 | memcpy(out, sctx, sizeof(*sctx)); |
171 | return 0; |
172 | } |
173 | |
174 | static int ppc_spe_sha256_import(struct shash_desc *desc, const void *in) |
175 | { |
176 | struct sha256_state *sctx = shash_desc_ctx(desc); |
177 | |
178 | memcpy(sctx, in, sizeof(*sctx)); |
179 | return 0; |
180 | } |
181 | |
182 | static struct shash_alg algs[2] = { { |
183 | .digestsize = SHA256_DIGEST_SIZE, |
184 | .init = sha256_base_init, |
185 | .update = ppc_spe_sha256_update, |
186 | .final = ppc_spe_sha256_final, |
187 | .export = ppc_spe_sha256_export, |
188 | .import = ppc_spe_sha256_import, |
189 | .descsize = sizeof(struct sha256_state), |
190 | .statesize = sizeof(struct sha256_state), |
191 | .base = { |
192 | .cra_name = "sha256" , |
193 | .cra_driver_name= "sha256-ppc-spe" , |
194 | .cra_priority = 300, |
195 | .cra_blocksize = SHA256_BLOCK_SIZE, |
196 | .cra_module = THIS_MODULE, |
197 | } |
198 | }, { |
199 | .digestsize = SHA224_DIGEST_SIZE, |
200 | .init = sha224_base_init, |
201 | .update = ppc_spe_sha256_update, |
202 | .final = ppc_spe_sha224_final, |
203 | .export = ppc_spe_sha256_export, |
204 | .import = ppc_spe_sha256_import, |
205 | .descsize = sizeof(struct sha256_state), |
206 | .statesize = sizeof(struct sha256_state), |
207 | .base = { |
208 | .cra_name = "sha224" , |
209 | .cra_driver_name= "sha224-ppc-spe" , |
210 | .cra_priority = 300, |
211 | .cra_blocksize = SHA224_BLOCK_SIZE, |
212 | .cra_module = THIS_MODULE, |
213 | } |
214 | } }; |
215 | |
216 | static int __init ppc_spe_sha256_mod_init(void) |
217 | { |
218 | return crypto_register_shashes(algs, ARRAY_SIZE(algs)); |
219 | } |
220 | |
221 | static void __exit ppc_spe_sha256_mod_fini(void) |
222 | { |
223 | crypto_unregister_shashes(algs, ARRAY_SIZE(algs)); |
224 | } |
225 | |
226 | module_init(ppc_spe_sha256_mod_init); |
227 | module_exit(ppc_spe_sha256_mod_fini); |
228 | |
229 | MODULE_LICENSE("GPL" ); |
230 | MODULE_DESCRIPTION("SHA-224 and SHA-256 Secure Hash Algorithm, SPE optimized" ); |
231 | |
232 | MODULE_ALIAS_CRYPTO("sha224" ); |
233 | MODULE_ALIAS_CRYPTO("sha224-ppc-spe" ); |
234 | MODULE_ALIAS_CRYPTO("sha256" ); |
235 | MODULE_ALIAS_CRYPTO("sha256-ppc-spe" ); |
236 | |